Title: Revisiting The Analysis of the Condition Of Streams In The Primary Region Of Mountaintop Mining/Valley Fill (MTM/VF) Coal Mining
1Revisiting The Analysis of the Condition Of
Streams In The Primary Region Of Mountaintop
Mining/Valley Fill (MTM/VF) Coal Mining
G. Pond and M. Passmore (USEPA Region 3)
MULTIVARIATE ANALYSIS OF TAXANOMIC DATA WITH
CHEMICAL AND PHYSICAL HABITAT VARIABLES Principal
Components Analysis (PCA) was used to explore the
multivariate character of a subset of the water
chemistry variables at the sites. Water quality
variables dominated by non-detects or with little
variation were not included in this analysis.
The PCA graph indicates the sites in
multidimensional space so that the longest axis
(the axis with the most variance) is the first
PCA axis, and the second longest axis is the
second PCA axis, perpendicular to the first. The
first few PCA axes indicate the greatest amount
of variation in the dataset and should contain
some significant patterns. In this case, the
first axis explained 71 of the variance in the
sites, and the second axis only explained an
additional 12 of the variance. In the EIS
dataset, potassium, selenium, sulfate, hardness,
alkalinity, conductivity, and sodium all had high
positive component loadings on axis 1. Note that
the filled sites are clearly separated from the
unmined sites along axis 1. The mined sites
plotted closer to the unmined sites in this
multivariate space, indicating their water
quality is more similar to the unmined
sites. Canonical Correspondence Analysis (CCA)
was used to relate the biotic variables (genera)
to abiotic variables (RBP habitat and median
water chemistry values). This analysis is a
multivariate, direct-gradient analysis method.
The axes of the final ordination are restricted
to be linear combinations of the environmental
variables and the taxa data. In this CCA, the
first two axes accounted for 42 of the variation
in the data, the first was 26 and the second was
16. The results of the CCA are presented with
the environmental variables plotted as arrows and
the taxa (called variables) and sites (called
cases) plotted as points in 2-dimensional space.
The sites are identified as mined, unmined and
filled by symbol. Each site lies at the centroid
of the points for species that occur in those
samples. The arrows indicate the direction of
maximum change for that environmental variable
and the length of the arrow is proportional to
the rate of change. The case (or site) axis
scores can be correlated to the original
environmental variables to further quantify the
relationship between where the site is positioned
and the individual variables. In the CCA case
diagram, the filled sites clearly contain very
different taxa from the unmined sites and the
arrows indicate several chemical or habitat
variables associated with those taxa differences.
Filled sites tended to have worse water quality
and habitat. In the CCA variable diagram, the
taxa associated with those chemical-physical
gradients are shown. The taxa associated with
the filled space are clearly more tolerant taxa
than those occupying the unmined space. This
ordination also indicates the lack of mayfly taxa
in the space associated with mine effluent or the
filled space. This space is dominated by
tolerant caddisflies (Hydropsyche and
Cheumatopsyche) and midge taxa (e.g. Cricotopus).
BACKGROUND In 1999 and 2000, EPA R3 characterized
and compared the ecological condition of unmined,
valley-filled, mined, residential and
filled-residential streams in the MTM/VF coal
fields of southern WV for a programmatic EIS
using a family-level Stream Condition Index
(WVSCI). Since that time, EPA has worked with
WVDEP to develop a genus-level index (GLIMPSS).
The vouchered EIS macroinvertebrate samples were
re-identified to genus level and reanalyzed using
the GLIMPSS. Relationships between the
ecological condition and various physical,
chemical and watershed characteristics were
examined using descriptive and multivariate
statistics. Results are shown here for the
filled, mined and unmined sites. Note that the
mined sites have some mining, but no valley
fills. The amount of mining in these watersheds
tends to be very small as most large scale
surface mining is associated with valley fills.
The genus-level index offers many refinements
over the family-level index, which are summarized
in table 1. These refinements offer a more
sensitive, and therefore accurate,
characterization of stream condition and causes
of impairment.
Table 1. Refinements offered by the genus-leve
GLIMPSS
These scatter plots show the relationship between
the GLIMPSS scores and mining in the watershed
and field conductivity for the Spring of 2000.
There is a strong and positive correlation
between the GLIMPSS scores and mining and
conductivity. This is similar to what was found
with the WVSCI.